Keynote by Helmuts Bems | CEO, Sonarworks
This keynote presents research-driven insights to ignite industry discussions on AI in music production.
The context
There were multiple reports in 2024 claiming that Spotify is increasingly serving AI-generated content to its user base. Here is just one example: https://www.fastcompany.com/91170296/spotify-ai-music. Is this overblown clickbait or just the tip of the iceberg of a bigger change coming? I will try to provide a framework for how to think.
In 2022, we formed a think tank at Sonarworks to try to understand the impact AI is going to have on the music industry. Since our first brainstorming session, we have conducted more than 100 interviews with industry professionals and consumers alike. Many ideas in this article came from various people we interviewed, from CEOs of industry companies to Grammy-winning engineers, artists, and casual music listeners.
The main focus is music production; however, we also touch on broader industry questions. The goal of this article is to initiate discussion and gather feedback. Nobody really knows what the future holds, but I strongly believe that diverse opinions and discussions will help everyone be better prepared for what’s to come.
Before diving into the details, let’s take a step back and look at the big picture of the music industry. As a proxy, I use this graph that most of you have probably seen.
U.S Recorded Music Revenues by Format
This is just a proxy since recording formats are only one aspect of this very complex industry. However, there is one obvious takeaway: The music industry has experienced massive revenue disruptions every 10 years. For example, the decline in CD sales took 10 years and reduced overall industry revenue from $26B to $8B.
From a production perspective, this is the start of the funnel. All these revenues trickle down into various complex revenue streams for artists, labels, retailers, producers, recording engineers, and product manufacturers. I joined the industry in 2012, so I do not have direct experience of how the previous downturn felt. The point is: Historically speaking, the music industry is both violent and cyclical – more so than any other industry I know.
It is not difficult to imagine that another major disruption is just around the corner. At first glance, AI meets all the criteria to be the next big disruptor. So far, AI has shattered all records in the speed of disruption. Just take a look at the AI adoption rates in the graph below:
AI Growth is Unstoppable
Spotify’s growth was 35% faster than Google’s. TikTok grew three times faster than Instagram, and ChatGPT outpaced TikTok by a factor of five. It took ChatGPT just two months to reach 100 million users. By now, we’ve become so accustomed to it that it feels normal, but historically, it is not. AI is the fastest-growing technology in human history by an exceptionally large margin. When I mentally combine AI technology with the music industry, I see many scenarios for highly disruptive change ahead.
AI could potentially disrupt the following aspects of the industry:
- Production process (the core focus of this article): how music is composed, recorded, mixed, and mastered.
- Production and playback products: the software and hardware used to produce and play back music.
- Music distribution: how end users access and consume content.
- Live shows: how live performances are set up and the content of the performance.
- Copyrights: how the money flows are distributed among various industry stakeholders.
AI Framework for Music Production
So far, the standard way of creating content has been entirely based on human input. Somebody comes up with an idea and writes it down. Then, some people come together and play the song for the purpose of recording it. After that, somebody mixes and masters the recording, and it gets distributed to the end listener. This is a human-led mode of music production. Let’s call it:
- Human-generated.
We believe that in the era of AI, there will be three new modes: - AI-assisted: Humans are involved in most music production stages, assisted by various AI-based tools. For instance, AI mixing or mastering tools, AI-based sound effects, or AI-based instrument and vocal transformers. In this mode, AI is present, but it is still humans steering the wheel.
- AI-generated: In this mode, the full content is generated via a prompt as input. In essence, the musical output is fully produced by AI, with only the initial prompt generated by the human. We believe prompt-edited content could also be a subset of this mode. In a prompt-edited manifestation, somebody (this could be the end listener) can take ready-made content and edit it via a prompt, thus creating a new content piece.
- Real-time AI-generated content: In this mode, content is fully generated by AI engines. However, unlike in mode 3, the input prompt is not generated by humans, but instead most likely automatically based on context. To be more specific, this could be your future music service generating new content for you specifically based on your mood or your environment. Note that I avoided calling this “streaming” because this mode can work from the cloud as well as locally. The best analogy for this is modern computer games that generate music based on events in the game.
Of course, the real world is not black and white, so the boundaries between these four modes will be blurred in real life. Still, this framework allows us to model how the future could take shape. Hypothetically, I could think of a step even further, where brain implants generate emotions and feelings directly in the brain. However, let’s delimit from this fifth mode until singularity gets closer. If you see another mode of production, let us know by commenting.
Modeling future scenarios
Modeling the future is a risky business. As Mike Tyson said, “Everyone has a plan until they get punched in the face.” We wanted to start by modeling how the future could play out as different scenarios. We chose the percentage of content produced by each production mode as the main proxy to model how the industry is going to change. We asked industry insiders to assign their subjective weights for each year over the next 10 years. This exercise revealed three different scenarios: slow, fast, and no disruption.
In order to present these scenarios, we combined 1) Human and 2) AI-assisted under one “Human mode” and 3) AI-generated with 4) Real-time under the second “AI mode.”
- The base (slow) disruption scenario suggests that AI advancement continues slowly, and it takes roughly 10 years for AI-generated to catch up to human-generated content. This is the base scenario for how we believe the AI transition is going to happen.
The technology for the disruption is already here; however, we believe the social change will take longer, and fully AI-generated content is going to face serious legal backlash from existing rights holders. This will be a roadblock for AI advancement in music creation. Additionally, developing AI solutions that are fully transparent and legally sound will take time, considering the complexity of music rights.
In the base scenario, human content creation will still be the dominant mode in the short to mid-term; however, in 10 years, AI-generated content could overtake human-generated content. The presence of AI will likely change the microeconomics of content creation, driving much more efficiency from creators. Creators will be driven to produce more content at a lower cost. Ironically, the biggest help for creators will be AI-assisted tools that will expand both creativity and efficiency in content creation.
Here are the arguments for and against the base scenario:
2) The fast disruption scenario. Technology continues advancing at an increasing speed with no barriers or speed bumps.
Based on the AI content news about Spotify that I mentioned at the beginning, AI content is already here and being commercialized. In the fast scenario, AI content overtakes human-generated in 5 years. AI continues its incredibly fast development speed and quickly becomes better than humans at creativity and efficiency. Technology advances so quickly that just as content gets AI-generated, it quickly moves to real-time contextual generation. The key drivers that enable this scenario:
- AI-generated content is good enough for most listeners compared to human-generated content.
- Real-time generation becomes technologically feasible.
Here are the arguments for and against this scenario:
3) The no-disruption scenario. There is serious backlash, either technologically, socially, or legally, that derails AI-generated music.
In this scenario, AI-generated takes a niche share in the overall industry by dominating the ‘elevator music’ category, which is, eventually, done in real-time. Humans continue creating most of the commercially viable content.
Here are the arguments for and against this scenario:
Like it or not, AI-generated music is here. In 2023, one of our investors sent me a link to an AI music generator and asked me to evaluate it. Back then, it felt like music coming out of a plastic kid’s toy. My response was, “This has no role to play in the professional music industry.” Then, a year later, that company released the next version of the engine, and it was really good. I actually ran a test where I took 5 songs from my weekly suggested playlist on Spotify and created a new playlist with 5 AI-generated songs mixed in. I asked people to listen to the playlist. None of the consumers noticed that the 5 songs were AI-generated and some even ended up liking the AI-generated songs more. On the other hand, some of the mixing engineers and producers who work at Sonarworks did notice some artifacts in the AI songs and remained skeptical. At the end of 2024, the next version of the engine came out. One of the previously skeptical, critical listeners approached me after playing with the engine for a few days and said to me that he had been converted because he managed to create a song that made a person cry.
Impact of AI Production Modes on Industry Stakeholders
Let’s look at how the advancement of AI in music production is going to affect some main stakeholders:
Producers and Composers
We believe that producers and composers will be the big winners in the AI era. They have the creative ideas, and they are the ones who can curate good content from bad. AI is going to enable them to be far more productive and thus more creative. Producers and composers in the AI era will be able to deliver more content than ever, without depending on others to deliver their parts.
Labels
The currently ongoing court cases against generative AI companies have mostly been brought by labels. Labels are arguing that AI is trained on the IP they own. If labels prevail in maintaining royalties for AI-generated content, there will be many long-term business opportunities for them. However, if generative AI companies end up being able to use existing content for training without paying for that content, labels will face an existential threat.
Streaming Services
AI-generated music presents both opportunities and risks for streaming platforms. In the base scenario, streaming services will see a positive impact on their performance since AI will lower costs by decreasing royalty payments to labels. Streaming services will have vastly more content, enabling more personalization and user engagement. However, if the advancement of AI continues at the current pace, generative models could be run locally in real time without the need to stream content. This would pose an existential threat to the business model. In such a scenario, streamers would face competition from “big tech,” possibly competing on the technology side.
Live Shows
We believe AI will generally have a positive effect on live shows. In the AI era, live shows could become the central expression for human connection with the artists. AI-generated music could improve live shows by introducing new creative possibilities, immersive experiences, and cost-saving opportunities. In the future, AI could generate real-time, adaptive music tailored to audience reactions, creating dynamic performances that evolve based on crowd energy, biometric data, and various other factors. Additionally, AI could assist human artists by generating backing tracks, orchestrations, or even improvisational elements during live shows. On the risk side, live shows could increasingly be challenged by virtual concerts featuring AI-generated artists, allowing fans to experience unique performances that may transcend physical limitations.
Professional Musicians
AI could reduce demand for human musicians in commercial industries such as advertising, film scoring, and background music production. Independent musicians may struggle to compete with an overwhelming flood of AI-generated content on streaming platforms, potentially leading to lower royalties and reduced visibility.
Hobby Musicians
AI-generated music will, in the short term, benefit hobby musicians by making music creation more accessible, intuitive, and enjoyable. With AI-powered tools, beginners will be able to generate melodies, harmonies, and beats effortlessly, even without formal training or advanced technical skills. This democratization of music production will allow more people to experiment with songwriting, remixing, and composition, fostering creativity across a wider audience. AI can also serve as a personal collaborator, offering suggestions, filling in gaps in compositions, or even helping hobbyists learn new musical techniques. This will be beneficial for the music instrument industry, as fewer barriers will drive more demand for music instruments. On the flip side, the presence of advanced AI creativity might discourage humans from learning to play musical instruments since AI might seem out of reach. Why labor to do something if that can be created with a click of a button?
Studio Engineers
AI-powered tools can automate many technical processes, allowing engineers to work more efficiently. However, by doing so, AI will take away the core added value of sound engineers. Even though initially, some engineers will benefit from increased efficiency, overall, there will be less demand for human engineers. For AI-generated music, the role of studio engineers will likely evolve toward curating, refining, and adding the human touch that AI lacks. Those who embrace AI as a tool will find new opportunities.
Equipment Companies
AI tools will greatly expand the range of the solutions that equipment companies will be able to offer and will influence companies by shifting demand toward AI-driven production tools, while potentially reducing the need for some traditional hardware. The biggest opportunity for hardware products will be in new user segments as more consumers will be drawn into creating content. However, as AI-powered software takes on roles like mixing, mastering, and content generation/transformation, there may be less reliance on high-end studio gear. Since AI requires significant computing power, some hardware companies may be challenged by big tech.
Software Companies
AI-generated music will significantly impact professional music software companies by accelerating the development of AI-driven tools while reshaping traditional music production workflows. As AI becomes more capable of composing, arranging, and even mixing music, software companies will increasingly integrate AI-powered features into digital audio workstations (DAWs), virtual instruments, and audio plugins. We estimate AI-assisted tools could be a $150M/y market by 2030. This will come at a cost of traditional production software.
Publishers
AI-generated music will have a transformative impact on music publishers, presenting both opportunities and challenges. On one hand, AI can generate vast amounts of music quickly, allowing publishers to expand their catalogs and license compositions for films, advertisements, and streaming platforms with minimal production costs. AI-assisted tools could also help publishers analyze market trends, predict hit songs, and optimize music distribution strategies. However, the widespread availability of AI-generated tracks could lead to market oversaturation, making it harder for traditionally published works to stand out. Additionally, legal and copyright challenges surrounding AI-created compositions may complicate royalty distribution and intellectual property rights, potentially disrupting traditional publishing revenue models.
Conclusion
In a nutshell, here are the key projections regarding the future impact on the music industry:
- There will be an exponential boost in the amount of content produced, and the cost associated with production will decrease.
- Less revenue will be captured by the legacy music industry. Big tech could generate new music-related revenue, however, that money will not flow downstream to artists, labels, and creatives as it used to.
- The role of sound engineers may be challenged with more power in the hands of producers and composers who will be more self-sufficient. AI will boost the efficiency of those who will embrace it.
- Commercial session musicians will be affected most as more output will be generated instead of recorded.
- Music creation software will be dominated by AI tools and the demand for legacy software will decrease.
- Some traditional music/audio hardware demand will decrease. Speakers and headphones will continue to be needed as people will still need those to listen to AI-generated content. Overall, there will be less demand for fancy equipment in the AI era. Future production will be more a factor of processing power than traditional hardware characteristics.
- In the short term, AI assistance will drive more people to explore music creativity; however, in the long run, people might be discouraged from even trying.
- More interesting human expressions for live shows. There could be fewer music performances and more of a show; however, we believe live shows could be the biggest winner overall, as we see them as the main channel for human interaction.
I believe the AI era for music is here. It is already much bigger than most people dare to think. This year at NAMM 2025 (the biggest PRO audio show), we were surprised by how prevalent AI use is among the top-level professionals we talked to. It seems AI use is widespread, but because of social stigma, many pros are reluctant to publicly admit it.
As a company, Sonarworks is fully committed to providing AI-assisted tools for creators, with our first product, VoiceAI, focusing on vocal transformation: creating any output with a human voice as input. We believe in engaging with AI with the goal of making it a helpful tool for humans.
On a more philosophical note, I have had discussions on where music comes from. Is it a product of mechanical human skill and the computational power of the human brain, or maybe there’s something divine and transcendental that exists out there, and humans only act as manifestors when the time is right? If the answer is the latter, then I am confident that AI is not going to be able to fully touch that divine, and humans will still be the main manifestators of the music that can touch another person deeply. Music has been such a big part of my life, and because of it, I might be biased to believe that there will always be a level of depth in music creation that will only be accessible to humans.
Do you believe AI will eventually get better than humans at creating new music? Let us know your thoughts by commenting on this article. Your opinion, even if just agreeing or disagreeing, will help everyone be more prepared.
About Sonarworks
Sonarworks is an award-winning audio technology innovator, delivering a uniquely perfected sound experience to every music creator and listener. Its patented technologies are used in over 250,000 studios and endorsed by 75+ Grammy-winning engineers who record A-list artists such as Lady Gaga, Madonna, Rihanna, Adele, Coldplay, and many more. In an ever-evolving world of sound, Sonarworks continues to lead the way, ensuring that every note, beat, and lyric is heard in its purest form. Sonarworks’ technologies power a range of products designed for creators. Among the most popular are SoundID Reference, which delivers consistent reference sound across speakers and headphones, and SoundID VoiceAI, a studio-grade voice and instrument transformer plugin for DAWs that boosts productivity and helps creators stay focused on their creative work.
